Research On Tabu Search Algorithm Based Video Object Tracking

With the continuous development of modern science technology, computer visioncomputing has been used in more and more fields. Among all of these technologies incomputer vision, video object tracking is one of the most nascent researching fields. Itinvolves lots of advanced technologies, such as artificial intelligence, patternrecognition, image processing and so on. Video object tracking has a great field ofapplication, such as medicine, traffic, military and manufacture, and more and morefields will develop and research on video object tracking. Nowadays, the object of videoobject tracking has switching from scene of static video to dynamic video, andresearchers will focus on two difficult questions: how to extract the feature from theobject, and how to find the object’s exact place in the next frame. Although there arelots of effective tracking algorithm, but not all of them perform well in the dynamicvideo. In the field of computer vision processing, scholars hope to find a real-time androbust tracking algorithm in any case.In this paper, we propose a Tabu search based video object tracking algorithm, theadvantages of this algorithm are list as follows: first of all, we adopt the basic Tabusearch algorithm in the video object tracking. Tabu search algorithm is a heuristic searchalgorithm. Due to the accuracy of the selection on the tracking object, Tabu search canbe effective used in the video object tracking. Secondly, we introduced the method toextract object’s feature and the matching algorithm. We introduced the color feature tosearch object. According to different feature spaces, we proposed comentropy based andBhattacharyyabased matching algorithm respectively. In the third part, we modify Tabusearch’s parameters and conditions according to the tracking conditions. What’s more,we proposed two optimized methods to improve the tracking performance: framedifference and optimized selection in neighborhood. By using these two methods,algorithm performs more stable and accurate during the static video. At last, we makedifferent experiments in different environments, also we make some research on theTabu’s parameters. According to the results of experiments, the algorithm we proposedin this paper performs real-time and robust not only in the static video, but also adaptivein the dynamic video. Also, Tabu search algorithm can’t handle the occlusion problems well.